Calibration

Because there is only a single biological snapshot of the fishery, there are many ways of guessing of how much fish there is. They give wildly different estimates but they are all a prior equally valid. What we want to do is try and use as many as possible and see what general rules are valid throughout.

Calibration here involves modifying the fishery parameters (catchabilities) such that for each biological scenario we match landings and catch histograms from Peter’s data.
Currently I have calibrated 3 setups, which I think span a decent range on the “optimistic-pessimistic” axis.

  1. Steady State: today’s landings and distribution can be reproduced indefinitely (assumes fixed recruitment); it is in a way extremely optimistic
  2. Optimistic: uses Steve’s method + Ren’s recruitment + cMSY (optimistic) + boxcar with 3 years of spinup time.
  3. Precipice: Uses Steve’s method + Ren’s recruitment but zeroes first year bins and replaces them with a single recruitment pulse.

We can show for each scenario the distribution of simulated landings (orange rectangles) versus real landings (black line) for each species and each population, as well as the difference between histograms of catch from data (“extrapolated”) and from the simulation. They mostly all match.

There are other scenarios I am currently calibrating and I will add to the report as they are done.

Biological Noise

Different scenarios have very different policy recommendations, even when focusing exclusively on what effort control to propose.

For example, in the steady state case there is no real conservation reason to restrict effort, since more biomass in the sea does not mean any more recruits later on. However we can still justify policy on the basis of higher profitability since more biomass in the sea means higher CPUE which implies achieve the same landings in a shorter amount of time. It’s hard to clear all the noise out but basically there is a sweet spot around 130-140 days where you get a quick enough stock refill and still have enough days to make it consistently profitable in less than 10 years.

The optimistic scenario is different because now conserving biomass will result in larger biomass later on. However it also implies that inaction will be very expensive, even in the short run.

Precipice, as the name implies, is a lot more pessimistic. Basically no policy involves a 20% drop in profitability within one year followed by more years of misery (this is attenuated in the long run by people leaving the fishery).
There is no policy that involve increase profits, only decisive effort control that will return the fishery to current levels of profitability after a decade.

In a way this uncertainty is however not just unavoidable but a justification for an adaptive management approach. Let’s look at how some indicator evolves around time and update our policy suggestions as new information comes in.

SPR adaptation

There are three issues to resolve if we want to target SPR.

  1. SPR moves slowly, and a target of 40% would require many years of effort control
  2. SPR estimated from catch is a decent indicator of real SPR but not perfect and tracks real SPR only slowly
  3. SPR from catch is susceptible to large errors in the face of recruitment failures.

To look at problem number one, let’s just focus on the Optimistic scenario and in particular impose a 50 max days out at sea. Even 15 years in, the real SPR has “only” moved from 6% to something like 22%, implying many more years of emergency effort control.

The “real” SPR is of course unknown, what we get is an estimate from catch data. However this estimate (and all its components like maturity ratio or average length caught) moves even slower than the real SPR.
In the next plot I compare the real biological SPR (dashed-line) against estimated SPR by looking at catches (solid-line). It is clearly a decent indicator because the correlation is evident, but it tends to react slowly when the changes are large (in both directions).

Finally recruitment noise can be a problem. This is actually something that the FAO’s review on length-based assessment mentions citing this paper (which I haven’t read).

The basic problem is simple: if there is a terrible year for recruitment, you will have few young catches in the years that follows. This will push down CPUE but push up the ratio of mature catches which will in turn increase the SPR value we extrapolated from it. To give a POSEIDON example let’s focus on the Steady State scenario (since being in equilibrium makes the effect clear). I compare normal runs (red lines) against runs where in year 4 there is no recruitment (blue lines). In both cases there is a large jump in estimated SPR the following year. The jump is in the wrong direction (that is, recruitment failure leads to more optimistic view of the stock).

Market solution

We discussed market premiums as a way to incentivate fishers towards catching more mature fish. However this seems to me counter-productive after running POSEIDON assuming a price premium on malabaricus.
The mildest problem is that fishers will pocket the surplus without changing much their fishing patterns. However, if the premium is high enough fishers may stay active when they would have otherwise quit as biomass dropped. That is, with a price premium we just push the fishery towards a more overfished equilibrium.

This is best seen in the Optimistic scenario where with no effort control biomass bottoms out after 10 years and this is followed by a mild recovery due to many boats quitting after years of negative profits. If we put a premium on malabaricus however, the situation worsens as the biomass needs to drop deeper and for longer in order to push fishers out.
There is a limit here somewhere around 250% premium simply because I am modelling exit but not entry in the fishery, otherwise this negative effect would be even more pronounced.